trureturingGitHub MATHEMATICAL ATLAS

RESEARCH / Focused target

Barry's A331474 Hankel generating-function conjecture

Literature status: not recheckedRepository record: proved

RELEASED FOUNDATIONS / PROPOSED CONNECTION

Research connections

1 released anchors
All source anchors (1)

Problem

The question

OEIS A331473 defines, at offset zero,

s_n = Sum_{k=0..n} (-1)^(n-k) * binomial(2*k+2,k).

OEIS A331474 is its literal Hankel transform

H_n = det(s_(i+j))_(0<=i,j<=n)

and Paul Barry's January 17, 2020 FORMULA line conjectures the full identity

Sum_{n>=0} H_n*x^n = (1 + 3*x + 16*x^2 - 8*x^3 + 36*x^4 - 12*x^5 + x^6 - x^7) / (1 + 7*x^2 + x^4)^2.

The resolved statement is D5/S3/Constants/Moments/A331474HankelGeneratingFunction.a331474_hankel_generating_function. It is an equality for PowerSeries.mk H, so the left side is the actual determinant sequence, not a recurrence-defined surrogate. The denominator has constant coefficient 1 and the formal right side uses its unit inverse.

Motivation

Our foothold

The displayed coefficients begin 1, 3, 2, -50, -43, 535, 487, -4983, -4654, 43174 and have alternating-sign blocks incompatible with a positive-real Gram argument. A proof therefore has to retain the literal moment source and control its signed Hankel determinants at every order.

Gap

Missing bridges

The exact source scope was pre-registered in issue #9676. A refresh on 2026-09-24 found A331474 still at revision 9, still marking Barry's formula (conjecture), and adding no proof reference. A331473 is revision 16 and still contains the literal alternating-binomial formula above.

The bounded source-family search included Bojičić-Petković-Barry (2025), Hacettepe Journal of Mathematics and Statistics 54, 1470-1478, DOI 10.15672/hujms.1564485. Its Theorems 3.1 and 5.1, after matching the first two source moments, predict a third moment of 9 or 10 rather than the literal value 12, so they do not subsume this source. An exact Crossref search for the A331474 identifier returned zero records. arXiv and OpenAlex searches were rate-limited with HTTP 429, leaving that coverage gap. The joeis direct Hankel generator and loda recurrence generator reproduce data but do not prove the bridge. These bounded readings do not establish worldwide absence, exhaustiveness, or priority.

Route

Proposed approach

The proof uses the pinned Mathlib Catalan power-series equation, polynomial basis-change determinants, matrix adjugates, Cramer identities, and formal power-series inversion. The source-specific bridge is D5/S3/Constants/Moments/A331474HankelBridge.literal_hankel_eq_signed_kernel.

  1. Differentiate the Catalan series to obtain the literal unsigned moments, then derive two period-two tails and their continuant error identity.
  2. Prove signed monomial and polynomial orthogonality for the monic continuants. The Gram diagonal entries are (-1)^k; negative norms are intentional.
  3. Use monic Gram diagonalization and a division-free adjugate identity. The actual adjacent-column conversion has coefficient c=-1, and Cramer's identity gives the signed finite kernel for every H_n.
  4. Derive the exact scalar recurrences for p and u, split by parity, and obtain the all-order recurrence for H. The values through H_9, including the n=8 and n=9 closure cases, follow from this bridge rather than a large finite determinant verification.
  5. Check every coefficient after multiplication by (1+7*x^2+x^4)^2, then multiply by its unit inverse to obtain the full generating-function identity.

Falsifier

What would falsify this route

Any natural index at which the determinant of the literal A331473 Hankel matrix differs from the corresponding coefficient of Barry's rational series would refute the claimed identity. A finite prefix agreement cannot prove the statement; the formal theorem quantifies over every coefficient through an identity of power series.

Evidence

Evidence to collect

  • Lean owners: D5/S3/Constants/Moments/A331474HankelBridge.lean and D5/S3/Constants/Moments/A331474HankelGeneratingFunction.lean.
  • The bridge theorem states the determinant kernel, both scalar initial-value sets, and both quantified recurrences in one public declaration. The final theorem states the complete identity for PowerSeries.mk H.

Triage

Scope assessment

theorem; resolution proved. Both public theorems have proof_shape: content: after all same-batch helpers are unfolded, the live chain contains the Catalan-tail continuant error, signed orthogonality, monic Gram diagonalization, division-free adjugate/Cramer conversion, scalar parity, and the coefficientwise all-order recurrence. The admission basis is escape-witness, and utility: none is accurate: these are general unbounded mathematical theorems, not finite-prefix certificates, bounded enumerations, checkers, or numeric reductions. The intermediate bridge receives no separate open-problem resolution claim; only the final generating-function theorem is the typed OpenProblemResolutionClaim(Proved) target.

ASSUMED-UNVERIFIED

Unverified assumptions

The arXiv and OpenAlex searches remain incomplete because their recorded requests returned HTTP 429. Crossref, joeis, loda, and the cited paper delimit the checked surfaces only. No global absence or worldwide priority claim is made. The formal result settles the literal source statement; it does not claim that every classical ingredient in its proof is new.

Dossier & anchor history

Loading release observations...